Detecção de Falhas em um Aplicativo Móvel Bancário

Abstract

The internet has changed the way banks deliver services to customers. Due to the high number of digital accesses, interruptions in Information Systems cause great damage to the financial system. One of the strategies implemented to achieve a stable environment is the continuous monitoring of services, as described by ITIL. Given the above, this work proposes a failure detection approach through data mining techniques using the CRISP-DM reference model. The approach involves evaluating data extracted from a web analytics tool, in real time, to identify critical failures in a mobile banking application. The effects of different feature engineering techniques, such as variable filtering, data standardization and synthetic sample generation, were evaluated in 7 classification algorithms. Finally, the results were compared, and the support vector machine was the one that obtained the best result, with an F1-Score of 0.954 and a ROC-AUC of 0.989.